Kinetic differences between the isoforms of glutamate decarboxylase: implications for the regulation of GABA synthesis

Glutamate decarboxylase (GAD) exists as two isoforms, GAD65 and GAD67. GAD activity is regulated by a cycle of activation and inactivation determined by the binding and release of its co-factor, pyridoxal 5'-phosphate. Holoenzyme (GAD with bound co-factor) decarboxylates glutamate to form GABA, but it also catalyzes a slower transamination reaction that produces inactive apoGAD (without bound co-factor). Apoenzyme can reassociate with pyridoxal phosphate to form holoGAD, thus completing the cycle. Within cells, GAD65 is largely apoenzyme (approximately 93%) while GAD67 is mainly holoenzyme (approximately 72%). We found striking kinetic differences between the GAD isoforms that appear to account for this difference in co-factor saturation. The glutamate dependent conversion of holoGAD65 to apoGAD was about 15 times faster than that of holoGAD67 at saturating glutamate. Aspartate and GABA also converted holoGAD65 to apoGAD at higher rates than they did holoGAD67. Nucleoside triphosphates (such as ATP) are known to affect the activation reactions of the cycle. ATP slowed the activation of GAD65 and markedly reduced its steady-state activity, but had little affect on the activation of GAD67 or its steady-state activity. Inorganic phosphate opposed the effect of ATP; it increased the rate of apoGAD65 activation but had little effect on apoGAD67 activation. We conclude that the apo-/holoenzyme cycle of inactivation and reactivation is more important in regulating the activity of GAD65 than of GAD67.     

Motifs and structural fold of the cofactor binding site of human glutamate decarboxylase.

The pyridoxal-P binding sites of the two isoforms of human glutamate decarboxylase (GAD65 and GAD67) were modeled by using PROBE (a recently developed algorithm for multiple sequence alignment and database searching) to align the primary sequence of GAD with pyridoxal-P binding proteins of known structure. GAD's cofactor binding site is particularly interesting because GAD activity in the brain is controlled in part by a regulated interconversion of the apo- and holoenzymes. PROBE identified six motifs shared by the two GADs and four proteins of known structure: bacterial ornithine decarboxylase, dialkylglycine decarboxylase, aspartate aminotransferase, and tyrosine phenol-lyase. Five of the motifs corresponded to the alpha/beta elements and loops that form most of the conserved fold of the pyridoxal-P binding cleft of the four enzymes of known structure; the sixth motif corresponded to a helical element of the small domain that closes when the substrate binds. Eight residues that interact with pyridoxal-P and a ninth residue that lies at the interface of the large and small domains were also identified. Eleven additional conserved residues were identified and their functions were evaluated by examining the proteins of known structure. The key residues that interact directly with pyridoxal-P were identical in ornithine decarboxylase and the two GADs, thus allowing us to make a specific structural prediction of the cofactor binding site of GAD. The strong conservation of the cofactor binding site in GAD indicates that the highly regulated transition between apo- and holoGAD is accomplished by modifications in this basic fold rather than through a novel folding pattern.     

Enzymatic characterization of a recombinant isoform hybrid of glutamic acid decarboxylase (rGAD67/65) expressed in yeast

BACKGROUND AND AIMS: Glutamic acid decarboxylase (GAD, EC 4.1.1.15) catalyses the conversion of glutamate to gamma-aminobutyric acid (GABA). The 65 kDa isoform, GAD65 is a potent autoantigen in type 1 diabetes, whereas GAD67 is not. A hybrid cDNA was created by fusing a human cDNA for amino acids 1-101 of GAD67 to a human cDNA for amino acids 96-585 of GAD65; the recombinant (r) protein was expressed in yeast and was shown to have equivalent immunoreactivity to mammalian brain GAD with diabetes sera. We here report on enzymatic and molecular properties of rGAD67/65. METHODS: Studies were performed on enzymatic activity of rGAD67/65 by production of 3H-GABA from 3H-glutamate, enzyme kinetics, binding to the enzyme cofactor pyridoxal phosphate (PLP), stability according to differences in pH, temperature and duration of storage, and antigenic reactivity with various GAD-specific antisera. RESULTS: The properties of rGAD67/65 were compared with published data for mammalian brain GAD (brackets). These included a specific enzyme activity of 22.7 (16.7) nKat, optimal pH for enzymatic activity 7.4 (6.8), K(m) of 1.3 (1.3) mM, efficient non-covalent binding to the cofactor PLP, and high autoantigenic potency. The stability of rGAD67/65 was optimal over 3 months at -80 degrees C, or in lyophilized form at -20 degrees C. CONCLUSIONS: Hybrid rGAD67/65 has enzymatic and other properties similar to those of the mixed isoforms of GAD in preparations from mammalian brain as described elsewhere, in addition to its previously described similar immunoreactivity.     

Effect of apocalmodulin on recombinant human brain glutamic acid decarboxylase

In this work, we report that the recombinant glutathione S-transferase (GST)-human L-glutamic acid decarboxylase (HGAD) isoforms, 65-kDa L-glutamic acid decarboxylase (GAD) (GST-HGAD65) fusion protein or free truncated HGAD65, were activated by apocalmodulin (ApoCaM) to an extent of 60%. Both truncated forms of GAD67 (tGAD67), HGAD67(Delta1-70) and HGAD67(Delta1-90), were markedly activated by ApoCaM to an extent of 141 and 85%, respectively, while GST-HGAD67 was not significantly affected. The activation appears to be due to an increase of GAD affinity for its cofactor, pyridoxal phosphate (PLP). This conclusion is based on the following observations. Firstly, the V(max) of GAD was increased when ApoCaM was present whereas the affinity for the substrate, glutamate, was not affected. Secondly, the affinity of GAD for PLP was increased in the presence of ApoCaM. Thirdly, results from calmodulin-agarose affinity column chromatography studies indicated a direct interaction or binding between ApoCaM and GAD. Fourthly, ApoCaM was found to be copurified with GAD65/GAD67 by anti-GAD65/67 immunoaffinity column using rat brain extract. Hence, it is proposed that a conformational change is induced when ApoCaM interacts with GAD65 or tGAD67, resulting in an increase of GAD affinity for PLP and the activation of GAD. The physiological significance of the interaction between GAD and ApoCaM is discussed.     

Cofactor and tryptophan accessibility and unfolding of brain glutamate decarboxylase

Cofactor and tryptophan accessibility of the 65-kDa form of rat brain glutamate decarboxylase (GAD) was investigated by fluorescence quenching measurements using acrylamide, I-, and Cs+ as the quenchers. Trp residues were partially exposed to solvent. I- was less able and Cs+ was more able to quench the fluorescence of Trp residues in the holoenzyme of GAD (holoGAD) than the apoenzyme (apoGAD). The fraction of exposed Trp residues were in the range of 30-49%. In contrast, pyridoxal-P bound to the active site of GAD was exposed to solvent. I- was more able and Cs+ was less able to quench the fluorescence of pyridoxal-P in holoGAD. The cofactor was present in a positively charged microenvironment, making it accessible for interactions with anions. A difference in the exposure of Trp residues and pyridoxal-P to these charged quenchers suggested that the exposed Trp residues were essentially located outside of the active site. Changes in the accessibility of Trp residues upon pyridoxal-P binding strongly supported a significant conformational change in GAD. Fluorescence intensity measurements were also carried out to investigate the unfolding of GAD using guanidine hydrochloride (GdnHCl) as the denaturant. At 0.8-1.5 M GdnHCl, an intermediate step was observed during the unfolding of GAD from the native to the denatured state, and was not found during the refolding of GAD from the denatured to native state, indicating that this intermediate step was not a reversible process. However, at >1.5 M GdnHCl for holoGAD and >2.0 M GdnHCl for apoGAD, the transition leading to the denatured state was reversible. It was suggested that the intermediate step involved the dissociation of native dimer of GAD into monomers and the change in the secondary structure of the protein. Circular dichroism revealed a decrease in the alpha-helix content of GAD from 36 to 28%. The unfolding pattern suggested that GAD may consist of at least two unfolding domains. Unfolding of the lower GdnHCl-resisting domain occurred at a similar concentration of denaturant for apoGAD and holoGAD, while unfolding of the higher GdnHCl-resisting domain occurred at a higher concentration of GdnHCl for apoGAD than holoGAD.     

Two Forms of the γ-Aminobutyric Acid Synthetic Enzyme Glutamate Decarboxylase Have Distinct Intraneuronal Distributions and Cofactor Interactions

Glutamate decarboxylase (GAD) catalyzes the production of gamma-aminobutyric acid (GABA), a major inhibitory neurotransmitter. The mammalian brain contains two forms of GAD, with Mrs of 67,000 and 65,000 (GAD67 and GAD65). Using a new antiserum specific for GAD67 and a monoclonal antibody specific for GAD65, we show that the two forms of GAD differ in their intraneuronal distributions: GAD67 is widely distributed throughout the neuron, whereas GAD65 lies primarily in axon terminals. In brain extracts, almost all GAD67 is in an active holoenzyme form, saturated with its cofactor, pyridoxal phosphate. In contrast, only about half of GAD65 (which is found in synaptic terminals) exists as active holoenzyme. We suggest that the relative levels of apo-GAD65 and holo-GAD65 in synaptic terminals may couple GABA production to neuronal activity.     

ATP's impact on the conformation and holoenzyme formation in relation to the regulation of brain glutamate decarboxylase

To investigate ATP as a potential factor in the regulation of brain glutamate decarboxylase (GAD), the impact of ATP on the enzyme conformation and holoenzyme formation was investigated. ATP at 100 microM quenches fluorescence emission intensity of the holoenzyme of GAD (holoGAD) by 18% after a correction for the inner filter effect and enhances fluorescence steady-state polarization from 0.158 to 0. 183 when excited at 280 or 295 nm. These findings suggest that ATP moderately changes the microenvironment of one or more tryptophan or tyrosine residues in holoGAD and alters these residues from a more mobile state to a less mobile one. A moderate ATP-induced conformational change in holoGAD is also supported by the observations that ATP increases the thermal denaturation temperature of holoGAD by 2 degrees C, as derived from temperature-dependent fluorescence spectra, and decreases the alpha-helical content of holoGAD by 8-10%, as determined by circular dichroism. Moreover, ATP does not affect the keto-enol tautomerization of holoGAD and has little or no direct effect on its activity, implying that the ATP interacting domain in holoGAD is not at the active site. Kinetics studies, as demonstrated by stopped-flow fluorescence and UV/visible spectroscopy, demonstrate that formation of holoGAD involves two steps: a fast reaction forming an apoGAD-cofactor intermediate complex, followed by a slow reaction involving the conformational change in the intermediate complex. ATP reduces the rate constant of the fast step to one-third and decreases the rate of the slow step and the intermediate complex formation constant to 60% of their original values. The present data suggest that ATP may regulate the interconversion between apoGAD and holoGAD by interacting with apoGAD rather than holoGAD. By slowing down the rate of intermediate complex formation, ATP reduces the amount of holoGAD formed.     

GABA production by glutamic acid decarboxylase is regulated by a dynamic catalytic loop

Gamma-aminobutyric acid (GABA) is synthesized by two isoforms of the pyridoxal 5'-phosphate-dependent enzyme glutamic acid decarboxylase (GAD65 and GAD67). GAD67 is constitutively active and is responsible for basal GABA production. In contrast, GAD65, an autoantigen in type I diabetes, is transiently activated in response to the demand for extra GABA in neurotransmission, and cycles between an active holo form and an inactive apo form. We have determined the crystal structures of N-terminal truncations of both GAD isoforms. The structure of GAD67 shows a tethered loop covering the active site, providing a catalytic environment that sustains GABA production. In contrast, the same catalytic loop is inherently mobile in GAD65. Kinetic studies suggest that mobility in the catalytic loop promotes a side reaction that results in cofactor release and GAD65 autoinactivation. These data reveal the molecular basis for regulation of GABA homeostasis.     

Quantum chemistry studies of the catalysis mechanism differences between the two isoforms of glutamic acid decarboxylase

The production of gamma-aminobutyric acid (GABA) is catalyzed by two isoforms of glutamic acid decarboxylase (GAD), using pyridoxal 5′-phosphate (PLP) as the cofactor. Between the two enzymes, GAD67 accounts for normal GABA requirement, while GAD65 stays inactive until emergent demand for GABA. Recent crystal structure findings revealed that the distinct conformation of a common catalytic loop of the enzymes may account for their different functions (Fenalti et al Nat Struct Mol Biol, 14:280-286, 2007). Enlightened by their inferences, we studied the underlying reaction mechanism of the two GAD isoforms using density functional theory (DFT). A rather complete reaction pathway is identified, including nine transition state (TS) structures and 14 intermediate (IM) structures. The rate limiting step occurs early during the reaction and involves a proton transfer. In the late stage, there are two pathways that involve C_4’ and C_α protonation by Tyr or Lys. Our calculations show that the reaction barriers corroborate the conjecture made by Fenalti et al.

Glutamate Decarboxylases in Nonneural Cells of Rat Testis and Oviduct: Differential Expression of GAD65 and GAD67

gamma-Aminobutyric acid (GABA) and its synthetic enzyme, glutamate decarboxylase (GAD), are not limited to the nervous system but are also found in nonneural tissues. The mammalian brain contains at least two forms of GAD (GAD67 and GAD65), which differ from each other in size, sequence, immunoreactivity, and their interaction with the cofactor pyridoxal 5'-phosphate (PLP). We used cDNAs and antibodies specific to GAD65 and GAD67 to study the molecular identity of GADs in peripheral tissues. We detected GAD and GAD mRNAs in rat oviduct and testis. In oviduct, the size of GAD, its response to PLP, its immunoreactivity, and its hybridization to specific RNA and DNA probes all indicate the specific expression of the GAD65 gene. In contrast, rat testis expresses the GAD67 gene. The GAD in these two reproductive tissues is not in neurons but in nonneural cells. The localization of brain GAD and GAD mRNAs in the mucosal epithelial cells of the oviduct and in spermatocytes and spermatids of the testis shows that GAD is not limited to neurons and that GABA may have functions other than neurotransmission.     

Regulation of GABA Metabolism in Discrete Rabbit Brain Regions under Methoxypyridoxine—Regional Differences in Cofactor Saturation and the Preictal Activation of Glutamate Decarboxylase Activity1

Abstract: The activities of the enzymes of the GABA system, glutamate decarboxylase (GAD) and GABA-transaminase, were measured in discrete regions of the rabbit brain before the onset and during the course of sustained epileptiform seizures induced by the vitamin B₆, analogue methoxypyridoxine (MP). GAD activities were measured in a reaction mixture alternatively containing the cofactor pyridoxal-5′-phosphate (PLP) in excess or containing no PLP (holoenzyme of GAD). A comparison between these two estimations showed that the apoenzyme of GAD is only partially saturated with cofactor and that the degree of saturation varied from brain area to brain area, being highest in cerebellar cortex and lowest in substantia nigra. Holoenzyme activity fell steeply after administration of 100 mg/kg MP. The regional degree of enzyme inhibition by MP was a function of the saturation of the apoenzyme with cofactor; i.e., a low rate of saturation resulted in a high degree of inhibition, and vice versa. That GAD from the regio inferior of the hippocampus did not fit into the scheme (strong inhibition is present although the degree of saturation is high) is discussed in view of the role of the hippocampus in seizure genesis and generalization. Inhibition of GAD activity by MP was completely reversible in vitro by excess PLP. Before the onset of seizures but not during their course, apoenzyme activity surpassed control levels. This preictal activation is significant in regio inferior of hippocampus, in superior colliculus, and in cerebellar cortex. GABA-transaminase activities were not significantly altered. The present study demonstrates that only investigation during the preictal period and in regional brain areas can reveal changes specific for the drug and perhaps representing the cause for seizure development, without being masked by additional alterations resulting from the severe functional and metabolic derangement during the ictal events. Thereby, it was disclosed that a decrease in vivo in the level of the enzyme product, GABA, is able to activate GAD.     

Regulation of γ-Aminobutyric Acid Synthesis in the Brain

Abstract: γ-Aminobutyric acid (GABA) is synthesized in brain in at least two compartments, commonly called the transmitter and metabolic compartments, and because reglatory processes must serve the physiologic function of each compartment, the regulation of GABA synthesis presents a complex problem. Brain contains at least two molecular forms of glutamate decarboxylase (GAD), the principal synthetic enzyme for GABA. Two forms, termed GAD₆₅ and GAD₆₇, are the products of two genes and differ in sequence, molecular weight, interaction with the cofactor, pyridoxal 5′-phosphate (pyridoxal-P), and level of expression among brain regions. GAD₆₅ appears to be localized in nerve terminals to a greater degree than GAD₆₇, which appears to be more uniformly distributed throughout the cell. The interaction of GAD with pyridoxal-P is a major factor in the short-term regulation of GAD activity. At least 50% of GAD is present in brain as apoenzyme (GAD without bound cofactor; apoGAD), which serves as a reservoir of inactive GAD that can be drawn on when additional GABA synthesis is needed. A substantial majority of apoGAD in brain is accounted for by GAD₆₅, but GAD₆₇ also contributes to the pool of apoGAD. The apparent localization of GAD₆₅ in nerve terminals and the large reserve of apo-GAD₆₅ suggest that GAD₆₅ is specialized to respond to short-term changes in demand for transmitter GABA. The levels of apoGAD and the holoenzyme of GAD (holoGAD) are controlled by a cycle of reactions that is regulated by physiologically relevant concentrations of ATP and other polyanions and by inorganic phosphate, and it appears possible that GAD activity is linked to neuronal activity through energy metabolism. GAD is not saturated by glutamate in synaptosomes or cortical slices, but there is no evidence that GABA synthesis in vivo is regulated physiologically by the availability of glutamate. GABA competitively inhibits GAD and converts holo- to apoGAD, but it is not clear if intracellular GABA levels are high enough to regulate GAD. There is no evidence of short-term regulation by second messengers. The syntheses of GAD₆₅ and GAD₆₇ proteins are regulated separately. GAD₆₇ regulation is complex; it not only is present as apoGAD₆₇, but the expression of GAD₆₇ protein is regulated by two mechanisms: (a) by control of mRNA levels and (b) at the level of translation or protein stability. The latter mechanism appears to be mediated by intracellular GABA levels.     

Structure of glutamate decarboxylase from E. coli: spectral studies

Structure of recombinant glutamate decarboxylase (GAD alpha) was studied by optical methods and electron microscopy. The active (pH 4.6) and inert (pH 6.3) holoGAD and apoGAD were investigated. Absorption and CD spectra were recorded in the range of 190 - 500 nm. Visible spectra were resolved into the bands corresponding to individual electron transitions using lognormal curves. The structures of predominant tautomers of internal aldimines were determined as ketoenamine at pH 4.6 and enolimine at pH 6.3. CD spectra show that holoGAD and apoGAD exhibit a negative band at 204 - 245 nm and a positive band near 190 - 204 nm. The contents of the secondary structure elements were estimated on the basis of the values of the mean residue ellipticity. Evidently, the main difference between the GAD forms studied is in the content of alpha-helix and random coil. HoloGAD has 50% of alpha-helix at pH 4.6 and 67% at pH 6.3, whereas apoGAD - 17 and 27%, respectively. Thus presented data establish the essential role of pyridoxal phosphate (PLP) in the organization of the GAD secondary structure due to tightening its polypeptide chain. It seems possible, that conformational changes induced by PLP binding stabilize the protein structure and promote the assembly of subunits into macromolecule, which was confirmed by electron microscopy.     

Transient increase in expression of a glutamate decarboxylase (GAD) mRNA during the postnatal development of the rat striatum.

We recently reported that the mammalian brain has two forms of the GABA synthetic enzyme glutamate decarboxylase (GAD, E.C. 4.1.1.15), which are the products of two genes. The two forms, which we call GAD65 and GAD67, differ from each other in sequence, molecular size, subcellular distribution, and interactions with the cofactor pyridoxal phosphate (PLP), with GAD65 activity more dependent than that of GAD67 on the continued presence of exogenous PLP. The existence of two GAD genes suggests that individual GABA neurons may be subject to differential regulation of GABA production. We have examined the expression of these two forms of GAD during postnatal development of the rat striatum to determine whether different classes of GABA neurons selectively express different amounts of the two GAD mRNAs. Here we present evidence for a dramatic developmental difference in the expression of the two mRNAs during postnatal development of the rat striatum. Using in situ hybridization to the two GAD mRNAs, we observed a selective increase in GAD65 mRNA during the second postnatal week, at the time when striatal matrix neurons innervate the substantia nigra (SN). PLP-dependent enzyme activity in the midbrain increases in parallel with increased expression of GAD65 mRNA in the striatum. We hypothesize that the innervation of the SN by striatal neurons triggers an increase in GAD65. The changing ratios of GAD65 and GAD67 in the striatum may contribute to the well-documented changes in seizure susceptibility that occur in early life.     

An analysis of the cross-reactivity of autoantibodies to GAD65 and GAD67 in diabetes

Background

Autoantibodies to GAD65 (anti-GAD65) are present in the sera of 70–80% of patients with type 1 diabetes (T1D), but antibodies to the structurally similar 67 kDa isoform GAD67 are rare. Antibodies to GAD67 may represent a cross-reactive population of anti-GAD65, but this has not been formally tested.

Methodology/Principal Findings

In this study we examined the frequency, levels and affinity of anti-GAD67 in diabetes sera that contained anti-GAD65, and compared the specificity of GAD65 and GAD67 reactivity. Anti-GAD65 and anti-GAD67 were measured by radioimmunoprecipitation (RIP) using ¹²⁵I labeled recombinant GAD65 and GAD67. For each antibody population, the specificity of the binding was measured by incubation with 100-fold excess of unlabeled GAD in homologous and heterologous inhibition assays, and the affinity of binding with GAD65 and GAD67 was measured in selected sera. Sera were also tested for reactivity to GAD65 and GAD67 by immunoblotting. Of the 85 sera that contained antibodies to GAD65, 28 contained anti–GAD67 measured by RIP. Inhibition with unlabeled GAD65 substantially or completely reduced antibody reactivity with both ¹²⁵I GAD65 and with ¹²⁵I GAD67. In contrast, unlabeled GAD67 reduced autoantibody reactivity with ¹²⁵I GAD67 but not with ¹²⁵I GAD65. Both populations of antibodies were of high affinity (>10¹⁰ l/mol).

Conclusions

Our findings show that autoantibodies to GAD67 represent a minor population of anti-GAD65 that are reactive with a cross-reactive epitope found also on GAD67. Experimental results confirm that GAD65 is the major autoantigen in T1D, and that GAD67 per se has very low immunogenicity. We discuss our findings in light of the known similarities between the structures of the GAD isoforms, in particular the location of a minor cross-reactive epitope that could be induced by epitope spreading.     

Histidine 282 in 5-aminolevulinate synthase affects substrate binding and catalysis

5-Aminolevulinate synthase (ALAS), the first enzyme of the heme biosynthetic pathway in mammalian cells, is a member of the alpha-oxoamine synthase family of pyridoxal 5'-phosphate (PLP)-dependent enzymes. In all structures of the enzymes of the -oxoamine synthase family, a conserved histidine hydrogen bonds with the phenolic oxygen of the PLP cofactor and may be significant for substrate binding, PLP positioning, and maintenance of the pKa of the imine nitrogen. In ALAS, replacing the equivalent histidine, H282, with alanine reduces the catalytic efficiency for glycine 450-fold and decreases the slow phase rate for glycine binding by 85%. The distribution of the absorbing 420 and 330 nm species was altered with an A420/A330 ratio increased from 0.45 to 1.05. This shift in species distribution was mirrored in the cofactor fluorescence and 300-500 nm circular dichroic spectra and likely reflects variation in the tautomer distribution of the holoenzyme. The 300-500 nm circular dichroism spectra of ALAS and H282A diverged in the presence of either glycine or aminolevulinate, indicating that the reorientation of the PLP cofactor upon external aldimine formation is impeded in H282A. Alterations were also observed in the K(Gly)d value and spectroscopic and kinetic properties, while the K(PLP)d increased 9-fold. Altogether, the results imply that H282 coordinates the movement of the pyridine ring with the reorganization of the active site hydrogen bond network and acts as a hydrogen bond donor to the phenolic oxygen to maintain the protonated Schiff base and enhance the electron sink function of the PLP cofactor.     

Regional differences in cofactor saturation of glutamate decarboxylase (GAD) in discrete brain nuclei of the rat

Glutamate decarboxylase (GAD) activities with and without added pyridoxal-5-phosphate were determined in discrete brain nuclei of freeze-dried samples. The distribution of GAD holoenzyme activity as well as the cofactor saturation, was found to be uneven in the discrete nuclei. In addition, it was found that repeated haloperidol treatment reduced GAD holoenzyme activity in the substantia nigra pars reticulata.     

ATP binding to brain l-glutamate decarboxylase: A study by affinity chromatography

The binding of ATP to brain l-glutamate decarboxylase (GAD) was studied by means of ATP-agarose chromatography, utilizing partially purified GAD from mouse brain after DEAE-cellulose chromatography and ammonium sulfate fractional precipitation. GAD was found to bind with a high affinity to the ATP-agarose with the ATP molecule linked to the beaded agarose through the N⁶-amino group. Agarose with ATP attached through the ribosyl hydroxyls was totally ineffective to bind the enzyme. GAD bound to the immobilized ATP could be dissociated by free ATP (10–50 mM), but not by ADP at a concentration as high as 100 mM. Mg²⁺ was not a required factor for the binding. The enzyme binding to the ATP-agarose occurred under a saturating concentration (50 μM) of pyridoxal 5′-phosphate (PLP). Moreover, GAD bound to the ATP-agarose was not dissociated by PLP even at 1.0 mM, indicating no competition of PLP with ATP for the same binding site on the enzyme. Kinetic characterization showed that binding of ATP raised the K_m of the enzyme for PLP. Our approach provides direct evidence that there is a specific binding site on GAD for ATP, which is distinct from the binding site for PLP.